OpenAI’s September 2026 release of GPT-6 Astra results, including an 82.9% score on the HLE-Diamond variant with tools, drove a sharp repricing in related markets by demonstrating capability well above the prior ~65% frontier on standard Humanity’s Last Exam. Current leaderboards show Anthropic’s Claude Opus 5.5 and Fable 5.1 variants ahead at 61–65% on full HLE evaluations, while OpenAI’s GPT-6 Astra sits near 54–55% on verified runs. Traders weigh OpenAI’s rapid iteration cadence and tool-augmented reasoning against Anthropic’s consistent top rankings and Google’s Gemini updates, with resolution hinging on any new OpenAI model submissions to the official leaderboard before December 31, 2026.